Introduction
What is a Knowledge Base?
A Knowledge Base in Flowable Design is a structured model that allows AI knowledge agents to retrieve, interpret, and respond using curated internal or external content. It transforms documents and other unstructured content into a semantically indexed format that supports intelligent querying within processes or via direct AI interaction.
By setting up a Knowledge Base, you enable your agents to generate more accurate and context-aware responses based on approved business knowledge - without hardcoding answers or writing complex logic.
Why Use a Knowledge Base?
Here are some key reasons to use the Knowledge Base feature in Flowable:
- AI-Powered Knowledge Retrieval: Agents can answer questions based on indexed internal content, reducing the need for manual data lookup or rule-based responses.
- Reusable Across Applications: Knowledge Bases can be used in processes, forms, or standalone search experiences.
- Scalable and Updatable: Supports structured updates via folder paths or file uploads, with automatic content extraction and indexing.
- Semantic Search Support: Content is processed and embedded in a vector store, enabling intelligent, context-aware retrieval via natural language.
Getting Started with Knowledge Base Configuration
To configure a Knowledge Base model, follow these steps in Flowable Design:
1. Select Type
| Type | Description |
|---|---|
| Process and Search | Content will be indexed to the Vector Store and querying is possible. |
| Search only | Only Vector Store search is available, the indexing must be done externally. |
2. Choose Input Datasource
| Datasource | Description |
|---|---|
| Content Items | Connects to a folder structure in Flowable's content repository. The Path to be indexed must be specified. |
| Static File | Upload one or multiple files directly into the Knowledge Base model. |
3. Define Content Processing Behavior
The Knowledge Base content goes through a processing pipeline before becoming searchable. You can configure the behavior at each stage:
Content Extraction
For content extraction the following options are available:
| Option | Description |
|---|---|
| Markdown | Extracts and formats content as Markdown. |
| Original Document | Only available for OpenAI Vector Store |
| Custom | Allows providing a custom Java implementation for extraction. |
Content Splitting
| Method | Description |
|---|---|
| Context-aware splitting | Uses semantic logic to split by topic or section. |
| Token text splitting | Splits based on token limits. |
| Disabled (No splitting) | Keeps content in full-length chunks. |
| Custom | Allows providing a custom Java implementation for splitting. |
Vector Store
| Option | Description |
|---|---|
| Elasticsearch | Default vector store for semantic search. |
| OpenAI Vector Store | Uses OpenAI's embedding and vector database capabilities. |
| Custom | Allows providing a custom Java implementation for storing vectors. |
Summary
Flowable's Knowledge Base model is your foundation for delivering intelligent, context-aware automation. This model transforms with the knowledge agent internal documentation into dynamic, searchable knowledge.